A safety helmets and overalls detection algorithm based on improved Yolov5

Workers in the factory need to wear safety helmets and overalls according to regulations. In order to detect whether someone violates the rules during working hours in real time, a deep learning detection algorithm based on yolov5 is proposed. The images are collected by the camera in the factory, and two different types of datasets are constructed. The experimental comparison shows that the detection accuracy and recall rate of the datasets have improved after adding additional categories. After optimizing the datasets, this paper adds coordinate attention mechanism to the backbone of yolov5 network. The results show that the improved algorithm has at least 1% improvement in recall and accuracy, and can effectively and timely detect illegal wear in complex operation scenarios.